The Person-Situation Debate

In 1968 Walter Mischel published a book that came close to demolishing personality psychology. His observation was straightforward: trait scores correlate with behaviour in any single situation at around 0.3. Most of the variation in what somebody does on a given occasion is therefore explained by something other than their disposition.

For about two decades that was read as showing that personality barely exists, and that situations do nearly all the work. It was a genuine crisis, and it produced a large literature arguing about whether traits were anything more than an illusion produced by observers who needed the world to be predictable.

The resolution, when it came, was almost embarrassingly simple. Seymour Epstein pointed out that traits are claims about distributions, not about occasions. Whether a conscientious person is punctual on a given Tuesday depends heavily on traffic, sleep and what else was happening. Whether they are punctual across fifty mornings is highly predictable. Aggregate behaviour across many occasions and trait correlations rise sharply: to 0.6, 0.7 and beyond, depending on how many occasions you average.

Mischel had measured single instances and found exactly what a trait theory should predict for single instances. Nobody had thought to state that expectation explicitly, so the finding arrived as a refutation rather than as a clarification.

The 0.3 also deserves some perspective. It is roughly the size of the relationship between many accepted medical treatments and their outcomes, and nobody concludes from that that the treatments do nothing. Effect sizes of that order look negligible when you imagine a single case and look substantial when you count the cases up. A modest correlation applied across many occasions produces large differences in accumulated outcomes, which is precisely how traits work in a life: not by determining what somebody does this afternoon, but by tilting thousands of afternoons very slightly in one direction until the sum of them is a visibly different life.

Mischel's own later position is more interesting than either side of the original argument. He proposed if-then behavioural signatures: somebody might be reliably aggressive when criticised by an authority figure and reliably calm when teased by a peer, and that profile is itself stable over years. The consistency is real; it lives at the level of situation-response pairs rather than in a single average across all situations. That is a more sophisticated notion of personality than the one his book attacked, and he is the person who produced it.

The practical application is uncomfortable and direct. A company that hires on one interview is making a single-instance judgement (one situation, one day, one interviewer) which is exactly the condition under which behaviour is least predictable from disposition. And then generalising from it enormously. Structured interviews help; work samples help more, because they aggregate across occasions and across occasions that are relevant. It is a very practical payoff from a debate that looked purely theoretical for twenty years.